
# General packages
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

# for scaling the data
from sklearn.preprocessing import StandardScaler 

# for clustering
from sklearn.cluster import KMeans 
from sklearn.metrics import silhouette_samples, silhouette_score

# Set display options to show all columns
pd.set_option('display.max_columns', None)
pd.set_option('display.width', None)

# Set the ggplot style 
plt.style.use('ggplot')

# Load the dataset
Data = pd.read_csv('Documents/Mastry_Project2/FinalSQL.csv')


# Excluding user_id column as it doesn't make sense to get its statistical summary
selected_columns = cohort_users.columns.difference(['user_id'])

# Generate the statistical summary for selected columns
cohort_users[selected_columns].describe().T


# flagging users based on flight bookings:True if the user has booked a flight and False if hasn't 
cohort_users['has_booked'] = ~cohort_users['cancellation_proportion'].isna()

# filling cancellation_proportion column null values with 0
cohort_users['cancellation_proportion'].fillna(0, inplace=True)


# Determine the position to insert the new column after the last_min_booking columns
position = cohort_users.columns.get_loc('avg_booking_departure_gap_days_hotels') + 1

# Insert the new column which is the maximum of the both columns at the desired position
cohort_users.insert(position, 'avg_booking_departure_gap_days', 
            cohort_users[['avg_booking_departure_gap_days_flights', 'avg_booking_departure_gap_days_hotels']].max(axis=1))

# Drop the original columns
cohort_users.drop(columns=['avg_booking_departure_gap_days_flights', 'avg_booking_departure_gap_days_hotels'], inplace=True)


# check the number of null values of the new column 'avg_booking_departure_gap_days'
null_values = cohort_users['avg_booking_departure_gap_days'].isna().sum()
print(f'We have {null_values} null values in avg_booking_departure_gap_days column.')

We have 432 null values in avg_booking_departure_gap_days column.


max_gap = cohort_users['avg_booking_departure_gap_days'].max()
cohort_users['avg_booking_departure_gap_days'].fillna(max_gap, inplace=True)


# Flag users who haven't booked a flight
cohort_users['has_booked_flight'] = ~cohort_users['avg_flight_price_usd'].isna()

# Columns to fill with 0
zero_fill_columns = ['avg_flight_discount_amount', 'avg_flight_seats', 'avg_checked_bags']

# Columns to fill with their medians
median_fill_columns = ['avg_flight_price_usd', 'avg_distance_flown_km']

# filling null values with zero
for col in zero_fill_columns:
    cohort_users[col].fillna(0, inplace=True)

# filling null values with median
for col in median_fill_columns:
    cohort_users[col].fillna(cohort_users[col].median(), inplace=True)


# Flag users who haven't booked a hotel
cohort_users['has_booked_hotel'] = ~cohort_users['avg_hotel_price_usd'].isna()

# Columns to fill with 0
zero_fill_columns = ['avg_hotel_discount_amount', 'avg_hotel_rooms', 'avg_stay_duration_day']

# Columns to fill with their medians
median_fill_columns = ['avg_hotel_price_usd']

# filling null values with zero
for col in zero_fill_columns:
    cohort_users[col].fillna(0, inplace=True)

# filling null values with median
for col in median_fill_columns:
    cohort_users[col].fillna(cohort_users[col].median(), inplace=True)

# checking the number of null values
null_values = cohort_users.isna().sum().sum()
print(f'We have {null_values} Null values in the dataset.')

We have 0 Null values in the dataset.


negative_numbers = cohort_users[cohort_users['avg_stay_duration_day'] < 0].shape[0]
percentage = round(cohort_users[cohort_users['avg_stay_duration_day'] < 0].shape[0]/
                   cohort_users.avg_stay_duration_day.count()*100,2)
print(f'Only {negative_numbers} records ({percentage}%) of avg_stay_duration_day column have negative amounts.')

Only 2 records (0.03%) of avg_stay_duration_day column have negative amounts.


# replacing the records with 0
cohort_users.loc[cohort_users['avg_stay_duration_day'] < 0, 'avg_stay_duration_day'] = 0

negative_numbers = cohort_users[cohort_users['avg_stay_duration_day'] < 0].shape[0]

# checking how many records we have with this criteria
print(f'Now {negative_numbers} records of the avg_stay_duration column is negative.')

Now 0 records of the avg_stay_duration column is negative.


# Select only numerical columns and excluding the user_id column
numerical_columns = [col for col in cohort_users.select_dtypes(include=['float64', 'int64']).columns if col != 'user_id']


# Set up the figure and axes
fig, axes = plt.subplots(len(numerical_columns), 2, figsize=(15, 4 * len(numerical_columns)))

for i, col in enumerate(numerical_columns):
    # Compute the outlier bounds: I commented out the lower_bound as in all the columns it's out of the range
    #lower_bound = cohort_users[col].mean() - 5 * cohort_users[col].std()
    upper_bound = cohort_users[col].mean() + 5 * cohort_users[col].std()
    mean = cohort_users[col].mean()
    
    # Plot histogram
    bins = int(np.log2(len(cohort_users[col])) + 1)  # Sturges' formula
    sns.histplot(cohort_users[col], ax=axes[i, 0], bins=bins, kde=True, edgecolor='black')
    axes[i, 0].set_title(f'Histogram of {col}')
    #axes[i, 0].axvline(lower_bound, color='r', linestyle='--', label=f'-5*STD')
    axes[i, 0].axvline(upper_bound, color='r', linestyle='--', label=f'5*STD')
    axes[i, 0].axvline(mean, color='r', linestyle='--', label='Mean')
    axes[i, 0].set_xlabel(col)
    axes[i, 0].set_ylabel('Frequency')
    axes[i, 0].legend()
    
    # Plot box plot
    sns.boxplot(x=cohort_users[col], ax=axes[i, 1], showfliers=True, boxprops=dict(alpha=0.7))
    axes[i, 1].axvline(upper_bound, color='r', linestyle='--', label=f'5*STD')  # Draw the 5*STD line on boxplot
    axes[i, 1].set_title(f'Boxplot of {col}')
    axes[i, 1].legend()

# Adjust layout
plt.tight_layout()
plt.show()


# we had 5998 records
number_of_outliers = 5998 - cohort_users.shape[0]
percentage = round(number_of_outliers/5998 * 100,2)
print(f'{number_of_outliers} ({percentage}%) records were outliers!')

329 (5.49%) records were outliers!


# Calculating the average_dollar_saved_per_kilometre metric 
avg_usd_saved_per_km = round(cohort_users['avg_flight_discount_amount'] 
                                                    * cohort_users['avg_flight_price_usd'] 
                                                    / cohort_users['avg_distance_flown_km'],2)

# Determine the position to insert the new column after the last_min_booking columns
position = cohort_users.columns.get_loc('avg_distance_flown_km') + 1

# Insert the new column which is the maximum of the both columns at the desired position
cohort_users.insert(position, 'avg_usd_saved_per_km', avg_usd_saved_per_km)


# List of columns to scale 
cols_to_scale = ['married', 'has_children', 'age', 'session_count', 'avg_session_duration_minute', 'avg_page_clicks', 
                 'total_trips', 'conversion_rate', 'weekend_trip_proportion', 'cancellation_proportion', 
                 'avg_booking_departure_gap_days', 'total_flights_booked', 'round_trips_proportion', 
                 'avg_flight_price_usd', 'avg_flight_discount_amount', 'discounted_flight_proportion', 'avg_flight_seats', 
                 'avg_checked_bags', 'avg_distance_flown_km', 'avg_usd_saved_per_km', 'total_hotels_booked', 
                 'avg_hotel_price_usd', 'avg_hotel_discount_amount', 'discounted_hotel_proportion', 
                 'avg_hotel_rooms', 'avg_stay_duration_day']

# Initialize the StandardScaler
scaler = StandardScaler()
    
# Apply scaling and create new columns with suffix "_scaled"
scaled_data = scaler.fit_transform(cohort_users[cols_to_scale])
scaled_columns = ["scaled_" + col for col in cols_to_scale]
cohort_users[scaled_columns] = scaled_data.round(2)

# check the dataset 
cohort_users.head()


# Extract the relevant columns for clustering
features = cohort_users[[
                  # Free hotel meal:
                  'scaled_has_children',
                  'scaled_age',
                  'scaled_avg_flight_seats', 
                  'scaled_avg_hotel_rooms',
                  # Free checked bag:
                  'scaled_avg_checked_bags',
                  'scaled_avg_stay_duration_day',
                  'scaled_avg_distance_flown_km',
                  'scaled_round_trips_proportion',
                  # No cancellation fees:
                  'scaled_cancellation_proportion',
                  'scaled_avg_booking_departure_gap_days',
                  # Exclusive discounts:
                  'scaled_conversion_rate',
                  'scaled_avg_page_clicks',
                  'scaled_avg_session_duration_minute',
                  'scaled_avg_flight_discount_amount',
                  'scaled_discounted_flight_proportion',
                  'scaled_avg_usd_saved_per_km',
                  'scaled_avg_hotel_discount_amount',
                  'scaled_discounted_hotel_proportion',  
                  # 1-night free hotel with a flight:
                  'scaled_total_flights_booked',
                  'scaled_total_hotels_booked',
                  'scaled_weekend_trip_proportion',
                  'scaled_total_trips'
                        ]]

# Compute the correlation matrix for the free_meal_index
correlation_matrix = features.corr()

# Set up the matplotlib figure
plt.figure(figsize=(15, 10))

# Generate a heatmap for the correlation matrix
sns.heatmap(correlation_matrix, annot=True, cmap="coolwarm", vmin=-1, vmax=1, linewidths=0.5, annot_kws={"size": 8})


# Adjust the x-axis labels for better readability
plt.xticks(rotation=90, ha='right')

# Set the title and show the plot
plt.title("Features Correlation Matrix Heatmap")
plt.tight_layout()
plt.show()


# Extract the relevant columns for clustering
features = cohort_users[[
                  # Free hotel meal:
                  #'scaled_has_children',
                  #'scaled_age',
                  #'scaled_avg_flight_seats', 
                  'scaled_avg_hotel_rooms',
                  # Free checked bag:
                  'scaled_avg_checked_bags',
                  #'scaled_avg_stay_duration_day',
                  #'scaled_avg_distance_flown_km',
                  #'scaled_round_trips_proportion',
                  # No cancellation fees:
                  'scaled_cancellation_proportion',
                  #'scaled_avg_booking_departure_gap_days',
                  # Exclusive discounts:
                  'scaled_conversion_rate',
                  #'scaled_avg_page_clicks',
                  #'scaled_avg_session_duration_minute',
                  #'scaled_avg_flight_discount_amount',
                  #'scaled_discounted_flight_proportion',
                  #'scaled_avg_usd_saved_per_km',
                  #'scaled_avg_hotel_discount_amount',
                  #'scaled_discounted_hotel_proportion',  
                  # 1-night free hotel with a flight:
                  #'scaled_total_flights_booked',
                  #'scaled_total_hotels_booked',
                  'scaled_weekend_trip_proportion',
                  #'scaled_total_trips'
                        ]]

# Display the first few rows of the selected features
features.head()


def get_silhouette_score(features, n_clusters):
    data_sample = features.sample(frac=1, random_state=1)
    model = KMeans(n_clusters=n_clusters, random_state=1)
    model.fit(data_sample)
    
    model_sil_score = silhouette_score(data_sample, model.labels_)
    
    print(f"The silhouette score using {n_clusters} number of clusters is: {round(model_sil_score, 3)}")
    
# calling the function for different number of clusters    
get_silhouette_score(features, 3)
get_silhouette_score(features, 4)
get_silhouette_score(features, 5)
get_silhouette_score(features, 6)
get_silhouette_score(features, 7)

The silhouette score using 3 number of clusters is: 0.314
The silhouette score using 4 number of clusters is: 0.367
The silhouette score using 5 number of clusters is: 0.351
The silhouette score using 6 number of clusters is: 0.378
The silhouette score using 7 number of clusters is: 0.394


# Retrieving the cluster labels assigned to each data point in our dataset.
cohort_users['cluster_label'] = model.labels_
model.labels_

array([2, 0, 3, ..., 4, 4, 1], dtype=int32)


cohort_users.head()


# Constructing a DataFrame to represent the cluster centroids with respective feature values.
model_centroids = pd.DataFrame(model.cluster_centers_, columns=features.columns)

# Displaying the coordinates/values of each cluster centroid in a tabular format.
display(model_centroids)

# Plotting a heatmap to visually represent the position and magnitude of each cluster centroid in the feature space.
plt.figure(figsize=(10, 6))
sns.heatmap(model_centroids.T, cmap="Blues", annot=True)
plt.xlabel("Cluster centroids", fontsize=14)
plt.tight_layout()
plt.title("Centroid positions in high-dimensional space", fontsize=15)

Text(0.5, 1.0, 'Centroid positions in high-dimensional space')


cluster_dict = {0:'Free Checked Bag',
               1: 'Exclusive Discounts',
               2: 'No Cancellation Fee',
               3: '1-night free hotel with a flight',
               4: 'Free Hotel Meal'} 

# Add a new column to the DataFrame to store the perk names
cohort_users['perk'] = cohort_users['cluster_label'].map(cluster_dict)

# Count the number of users in each segment
segment_counts = cohort_users['perk'].value_counts().reset_index()
segment_counts.columns = ['perk', 'user_count']

# Display the number of users in each segment
segment_counts


grouped_df = features.join(cohort_users['perk']).groupby('perk').mean()
grouped_df


filtered_columns = [col for col in grouped_df.columns if col != 'scaled_cancellation_proportion'] 

# Extract unique segment labels from the 'perk' column
segments = cohort_users['perk'].unique()

# Calculate the number of rows and columns for the subplot grid
n_rows = len(filtered_columns)  # One row per feature
n_cols = 2  # Two columns per row

# Create subplots with explicit layout
fig, axes = plt.subplots(n_rows, n_cols, figsize=(15, 4 * len(filtered_columns)))


# Loop through each feature and create KDE plots for each segment
for i, feature in enumerate(filtered_columns):
    row, col = divmod(i, n_cols)  # Calculate row and column for current subplot
    for segment in segments:
        sns.kdeplot(
            cohort_users[cohort_users['perk'] == segment][feature],
            label=segment,
            ax=axes[row, col]
        )
    
    axes[row, col].set_title(f'Distribution of {feature} Within Each Segment')
    axes[row, col].set_xlabel(feature)
    axes[row, col].set_ylabel('Density')
    axes[row, col].legend(title='Segment')


# Hide any remaining empty subplots
for i in range(len(filtered_columns), n_rows * n_cols):
    row, col = divmod(i, n_cols)  # Calculate row and column for current subplot
    fig.delaxes(axes[row, col])

plt.tight_layout()
plt.show()


# Calculate the number of rows for the subplot grid (one row per feature)
n_rows = len(features.columns)

# Create subplots with explicit layout, one subplot per row
fig, axes = plt.subplots(n_rows, 1, figsize=(12, 4 * len(features.columns)))

# Loop through each feature to create box plots for each segment
for i, feature in enumerate(features.columns):
    sns.boxplot(y='perk', x=feature, data=cohort_users, ax=axes[i])
    
    axes[i].set_title(f'Box Plot of {feature} Across Each Perk')
    axes[i].set_xlabel('Feature Values')
    axes[i].set_ylabel('Perk')

plt.tight_layout()
plt.show()


from sklearn.metrics.pairwise import euclidean_distances

# Make a copy of the DataFrame slice
features_copy = features.copy()

# Add the cluster labels to the DataFrame containing the features
features_copy['cluster_label'] = cohort_users['cluster_label']
features_copy['perk'] = cohort_users['perk']

# Initialize an empty column to store the distances to the cluster centroid
features_copy['dist_center'] = None

# Loop through each cluster to calculate the distance of each point to its cluster's centroid
for cluster in model_centroids.index:
    cluster_sample = features_copy.loc[features_copy['cluster_label'] == cluster]
    
    features_copy.loc[features_copy['cluster_label'] == cluster, 'dist_center'] = euclidean_distances(
        cluster_sample.iloc[:, :-3],  # Exclude the last three columns: 'cluster_label', 'perk', 'dist_center'
        model_centroids.loc[cluster, :].values.reshape(1, -1)
    )
    
# Calculate the average distance of the data points in each cluster to their respective centroid
avg_dist_to_centroid = features_copy.groupby('perk', as_index=False)['dist_center'].mean()
avg_dist_to_centroid


# Define the function to draw representative examples from each cluster (perk category in this case)
def draw_examples(df, perk, max_dist, n):
    print(f"Examples from cluster {perk} with less than {max_dist} distance to the cluster centroid")
    return df.loc[(df['perk'] == perk) & (df['dist_center'] < max_dist)].sample(n)

# Draw 5 representative examples from the 'No Cancellation Fee' cluster
# Replace 0.4 with the appropriate maximum distance based on your specific dataset
draw_examples(features_copy, perk='Exclusive Discounts', max_dist=0.4, n=5)

Examples from cluster Exclusive Discounts with less than 0.4 distance to the cluster centroid


# Calculate various metrics for each cluster (perk)
cluster_metrics = features_copy.groupby('perk', as_index=False).agg({
    'scaled_conversion_rate': 'mean',
    'scaled_avg_hotel_rooms': 'mean',
    'scaled_avg_checked_bags': 'mean',
    'scaled_cancellation_proportion': 'mean',
    'scaled_weekend_trip_proportion': 'mean'
})

# Rename columns for clarity
cluster_metrics.columns = ['Perk', 'Avg Conversion Rate', 'Avg Hotel Rooms', 'Avg Checked Bags', 
                           'Avg Cancellation Proportion', 'Avg Weekend Trip Proportion']

# Display the calculated metrics as a table
cluster_metrics


# Count the number of customers in each segment
segment_count = cohort_users['perk'].value_counts().reset_index()
segment_count.columns = ['Segment', 'Number of Customers']

# Plotting the bar chart
plt.figure(figsize=(12, 6))
sns.barplot(x='Number of Customers', y='Segment', data=segment_count, palette='cool')
plt.title('Distribution of Customers Across Different Segments')
plt.xlabel('Number of Customers')
plt.ylabel('Segment')

# Annotate bars with count values
for index, value in enumerate(segment_count['Number of Customers']):
    plt.text(value, index, str(value), ha='center', va='center', fontsize=10, color='black')

#plt.tight_layout()
plt.show()


# to use it in the presentation file:

def func(pct, allvalues): 
    absolute = int(pct/100.*np.sum(allvalues))
    return "{:.1f}%\n({:d} users)".format(pct, absolute)

# Data to plot
sizes = segment_count['Number of Customers']
labels = segment_count['Segment']
colors = sns.color_palette('cool', len(labels))

# Plotting the Pie chart
plt.figure(figsize=(7,7))
plt.pie(sizes, labels=labels, colors=colors, autopct=lambda pct: func(pct, sizes), startangle=140, pctdistance=0.85)

# Drawing center circle for 'Donut' style
centre_circle = plt.Circle((0,0),0.70,fc='white')
fig = plt.gcf()
fig.gca().add_artist(centre_circle)

# Use the y parameter to adjust the position of the title
plt.title('Distribution of Customers Across Different Segments', y=1.08)

# Equal aspect ratio ensures that pie is drawn as a circle
plt.axis('equal')  
plt.tight_layout()
plt.show()


# Calculate the average age for each segment
avg_age_segment = cohort_users.groupby('perk')['age'].mean().reset_index()
avg_age_segment.columns = ['Segment', 'Average Age']

# Plotting the bar chart for average age
plt.figure(figsize=(12, 6))
ax = sns.barplot(x='Average Age', y='Segment', data=avg_age_segment, palette='cool')

# Annotate the average age of each segment
for index, value in enumerate(avg_age_segment['Average Age']):
    ax.text(value - 0.5, index, str(round(value, 2)))

plt.title('Average Age of Customers Across Different Segments')
plt.xlabel('Average Age')
plt.ylabel('Segment')
plt.show()


# Gender analysis: drawing the countplot for gender distribution across different perks using
plt.figure(figsize=(12, 6))
sns.countplot(data=cohort_users, x='perk', hue='gender', palette="cool")
plt.title('Gender Distribution by Perk')
plt.xlabel('Perk')
plt.ylabel('Number of Customers')
plt.xticks(rotation=15)
plt.show()


# Plotting the distribution of customers across top 5 cities and their most preferred perk
top_5_cities = cohort_users['home_city'].value_counts().nlargest(5).index.tolist()

# Filter the data to only include the top 5 cities
top_5_cities_data = cohort_users[cohort_users['home_city'].isin(top_5_cities)]

# Generate a horizontal bar chart for the top 5 cities
plt.figure(figsize=(15, 8))
sns.countplot(data=top_5_cities_data, y='home_city', hue='perk', palette="cool")
plt.title('Top 5 Cities: Distribution of Customers by Preferred Perk')
plt.xlabel('Number of Customers')
plt.ylabel('City')
plt.legend(title='Preferred Perk', bbox_to_anchor=(1, 1), loc='upper left')
plt.tight_layout()
plt.show()


final_segmentation_df = cohort_users[['user_id','perk']]

# Save the DataFrame to a CSV file
final_segmentation_df.to_csv('TravelTide_Customer_Segmentation_Kmean.csv', index=False)

# Display the first few rows of the final DataFrame
final_segmentation_df.head()

	count	mean	std	min	25%	50%	75%	max
age	5998.0	41.014672	12.046715	16.00	34.0000	41.000	48.0000	88.00
avg_booking_departure_gap_days_flights	5206.0	21.818417	48.355668	2.35	6.6100	7.645	9.0200	365.08
avg_booking_departure_gap_days_hotels	5435.0	12.981943	27.731505	1.50	6.5800	7.730	9.0500	329.83
avg_checked_bags	5206.0	0.603338	0.512678	0.00	0.2500	0.500	1.0000	6.60
avg_distance_flown_km	5206.0	2348.176104	1749.255961	27.52	1476.5125	2000.040	2696.1900	15954.76
avg_flight_discount_amount	4721.0	0.141010	0.070932	0.05	0.1000	0.130	0.1800	0.55
avg_flight_price_usd	5206.0	550.205344	739.590156	5.35	282.7750	392.795	565.9625	14280.38
avg_flight_seats	5206.0	1.222578	0.454272	1.00	1.0000	1.000	1.3300	6.60
avg_hotel_discount_amount	4063.0	0.112437	0.054025	0.05	0.0800	0.100	0.1500	0.45
avg_hotel_price_usd	5435.0	178.272981	84.673031	24.00	124.3300	163.000	211.8350	1063.00
avg_hotel_rooms	5435.0	1.198267	0.352250	1.00	1.0000	1.000	1.3300	4.00
avg_page_clicks	5998.0	18.782508	11.969869	4.13	12.5600	15.670	19.8150	163.90
avg_session_duration_minute	5998.0	1.780123	1.602640	0.00	1.1000	1.440	1.8900	16.40
avg_stay_duration_day	5435.0	3.804942	3.007981	-1.00	2.0000	3.000	5.0000	30.00
cancellation_proportion	5566.0	0.047528	0.160093	0.00	0.0000	0.000	0.0000	1.00
conversion_rate	5998.0	0.333846	0.185690	0.00	0.2200	0.330	0.5000	1.00
discounted_flight_proportion	5998.0	0.187049	0.136935	0.00	0.1100	0.130	0.2500	0.70
discounted_hotel_proportion	5998.0	0.145782	0.126332	0.00	0.0000	0.130	0.2500	0.75
round_trips_proportion	5998.0	0.827904	0.351102	0.00	1.0000	1.000	1.0000	1.00
session_count	5998.0	8.427309	0.790536	8.00	8.0000	8.000	9.0000	14.00
total_flights_booked	5998.0	2.397966	1.577156	0.00	1.0000	2.000	3.0000	9.00
total_hotels_booked	5998.0	2.527509	1.511832	0.00	1.0000	2.000	4.0000	8.00
total_trips	5998.0	2.805435	1.589629	0.00	2.0000	3.000	4.0000	9.00
weekend_trip_proportion	5998.0	0.049298	0.137947	0.00	0.0000	0.000	0.0000	1.00

	user_id	sign_up_date	age	gender	married	has_children	home_country	home_city	session_count	avg_session_duration_minute	avg_page_clicks	total_trips	conversion_rate	weekend_trip_proportion	cancellation_proportion	avg_booking_departure_gap_days	total_flights_booked	round_trips_proportion	avg_flight_price_usd	avg_flight_discount_amount	discounted_flight_proportion	avg_flight_seats	avg_checked_bags	avg_distance_flown_km	avg_usd_saved_per_km	total_hotels_booked	avg_hotel_price_usd	avg_hotel_discount_amount	discounted_hotel_proportion	avg_hotel_rooms	avg_stay_duration_day	has_booked	has_booked_flight	has_booked_hotel	scaled_married	scaled_has_children	scaled_age	scaled_session_count	scaled_avg_session_duration_minute	scaled_avg_page_clicks	scaled_total_trips	scaled_conversion_rate	scaled_weekend_trip_proportion	scaled_cancellation_proportion	scaled_avg_booking_departure_gap_days	scaled_total_flights_booked	scaled_round_trips_proportion	scaled_avg_flight_price_usd	scaled_avg_flight_discount_amount	scaled_discounted_flight_proportion	scaled_avg_flight_seats	scaled_avg_checked_bags	scaled_avg_distance_flown_km	scaled_avg_usd_saved_per_km	scaled_total_hotels_booked	scaled_avg_hotel_price_usd	scaled_avg_hotel_discount_amount	scaled_discounted_hotel_proportion	scaled_avg_hotel_rooms	scaled_avg_stay_duration_day
0	23557	2021-07-22	64	F	True	False	usa	new york	12	6.93	17.50	4	0.33	0.00	0.25	248.05	2	1.0	518.29	0.15	0.21	1.00	1.00	2680.73	0.03	4	157.33	0.18	0.29	1.33	7.33	True	True	True	1.13	-0.69	1.93	4.81	4.98	-0.04	0.73	-0.05	-0.38	2.59	2.07	-0.27	0.5	0.15	0.46	0.21	-0.06	1.10	0.45	0.31	0.95	-0.24	1.52	1.22	0.52	1.44
1	94883	2022-02-07	51	F	True	False	usa	kansas city	12	0.50	8.33	3	0.25	0.00	0.00	60.98	3	1.0	1784.95	0.10	0.08	1.67	1.00	5057.35	0.04	2	90.00	0.08	0.17	1.50	0.50	True	True	True	1.13	-0.69	0.83	4.81	-1.06	-0.92	0.10	-0.48	-0.38	-0.26	0.18	0.36	0.5	3.62	-0.13	-0.75	1.23	1.10	2.51	0.79	-0.37	-1.16	0.06	0.24	0.89	-1.03
3	101961	2022-02-17	43	F	True	False	usa	boston	12	2.00	18.17	7	0.58	0.14	0.00	7.34	6	1.0	321.53	0.13	0.25	1.00	0.50	1753.36	0.02	7	150.29	0.10	0.08	1.00	3.14	True	True	True	1.13	-0.69	0.15	4.81	0.35	0.03	2.61	1.29	0.82	-0.26	-0.36	2.26	0.5	-0.39	0.22	0.51	-0.06	-0.00	-0.35	-0.17	2.94	-0.34	0.35	-0.49	-0.19	-0.08
9	149058	2022-04-14	50	F	False	True	usa	birmingham	12	2.79	15.36	6	0.50	0.00	0.17	58.79	6	1.0	664.40	0.05	0.21	1.56	0.67	2304.81	0.01	6	165.67	0.00	0.14	1.11	8.56	True	True	True	-0.89	1.45	0.75	4.81	1.09	-0.24	1.99	0.86	-0.38	1.68	0.16	2.26	0.5	0.55	-0.72	0.21	1.01	0.37	0.13	-0.65	2.28	-0.13	-1.11	0.00	0.05	1.88
10	152583	2022-04-17	37	F	False	False	usa	colorado springs	10	6.17	21.25	2	0.20	0.00	0.50	181.16	2	1.0	3769.43	0.15	0.25	1.80	1.20	10253.52	0.06	2	213.00	0.15	0.42	1.00	2.00	True	True	True	-0.89	-0.69	-0.35	2.13	4.27	0.33	-0.53	-0.75	-0.38	5.43	1.40	-0.27	0.5	9.06	0.46	0.51	1.47	1.54	6.99	1.76	-0.37	0.52	1.08	2.27	-0.19	-0.49

	scaled_avg_hotel_rooms	scaled_avg_checked_bags	scaled_cancellation_proportion	scaled_conversion_rate	scaled_weekend_trip_proportion
0	0.52	1.10	2.59	-0.05	-0.38
1	0.89	1.10	-0.26	-0.48	-0.38
3	-0.19	-0.00	-0.26	1.29	0.82
9	0.05	0.37	1.68	0.86	-0.38
10	-0.19	1.54	5.43	-0.75	-0.38

	user_id	sign_up_date	age	gender	married	has_children	home_country	home_city	session_count	avg_session_duration_minute	avg_page_clicks	total_trips	conversion_rate	weekend_trip_proportion	cancellation_proportion	avg_booking_departure_gap_days	total_flights_booked	round_trips_proportion	avg_flight_price_usd	avg_flight_discount_amount	discounted_flight_proportion	avg_flight_seats	avg_checked_bags	avg_distance_flown_km	avg_usd_saved_per_km	total_hotels_booked	avg_hotel_price_usd	avg_hotel_discount_amount	discounted_hotel_proportion	avg_hotel_rooms	avg_stay_duration_day	has_booked	has_booked_flight	has_booked_hotel	scaled_married	scaled_has_children	scaled_age	scaled_session_count	scaled_avg_session_duration_minute	scaled_avg_page_clicks	scaled_total_trips	scaled_conversion_rate	scaled_weekend_trip_proportion	scaled_cancellation_proportion	scaled_avg_booking_departure_gap_days	scaled_total_flights_booked	scaled_round_trips_proportion	scaled_avg_flight_price_usd	scaled_avg_flight_discount_amount	scaled_discounted_flight_proportion	scaled_avg_flight_seats	scaled_avg_checked_bags	scaled_avg_distance_flown_km	scaled_avg_usd_saved_per_km	scaled_total_hotels_booked	scaled_avg_hotel_price_usd	scaled_avg_hotel_discount_amount	scaled_discounted_hotel_proportion	scaled_avg_hotel_rooms	scaled_avg_stay_duration_day	cluster_label
0	23557	2021-07-22	64	F	True	False	usa	new york	12	6.93	17.50	4	0.33	0.00	0.25	248.05	2	1.0	518.29	0.15	0.21	1.00	1.00	2680.73	0.03	4	157.33	0.18	0.29	1.33	7.33	True	True	True	1.13	-0.69	1.93	4.81	4.98	-0.04	0.73	-0.05	-0.38	2.59	2.07	-0.27	0.5	0.15	0.46	0.21	-0.06	1.10	0.45	0.31	0.95	-0.24	1.52	1.22	0.52	1.44	2
1	94883	2022-02-07	51	F	True	False	usa	kansas city	12	0.50	8.33	3	0.25	0.00	0.00	60.98	3	1.0	1784.95	0.10	0.08	1.67	1.00	5057.35	0.04	2	90.00	0.08	0.17	1.50	0.50	True	True	True	1.13	-0.69	0.83	4.81	-1.06	-0.92	0.10	-0.48	-0.38	-0.26	0.18	0.36	0.5	3.62	-0.13	-0.75	1.23	1.10	2.51	0.79	-0.37	-1.16	0.06	0.24	0.89	-1.03	0
3	101961	2022-02-17	43	F	True	False	usa	boston	12	2.00	18.17	7	0.58	0.14	0.00	7.34	6	1.0	321.53	0.13	0.25	1.00	0.50	1753.36	0.02	7	150.29	0.10	0.08	1.00	3.14	True	True	True	1.13	-0.69	0.15	4.81	0.35	0.03	2.61	1.29	0.82	-0.26	-0.36	2.26	0.5	-0.39	0.22	0.51	-0.06	-0.00	-0.35	-0.17	2.94	-0.34	0.35	-0.49	-0.19	-0.08	3
9	149058	2022-04-14	50	F	False	True	usa	birmingham	12	2.79	15.36	6	0.50	0.00	0.17	58.79	6	1.0	664.40	0.05	0.21	1.56	0.67	2304.81	0.01	6	165.67	0.00	0.14	1.11	8.56	True	True	True	-0.89	1.45	0.75	4.81	1.09	-0.24	1.99	0.86	-0.38	1.68	0.16	2.26	0.5	0.55	-0.72	0.21	1.01	0.37	0.13	-0.65	2.28	-0.13	-1.11	0.00	0.05	1.88	2
10	152583	2022-04-17	37	F	False	False	usa	colorado springs	10	6.17	21.25	2	0.20	0.00	0.50	181.16	2	1.0	3769.43	0.15	0.25	1.80	1.20	10253.52	0.06	2	213.00	0.15	0.42	1.00	2.00	True	True	True	-0.89	-0.69	-0.35	2.13	4.27	0.33	-0.53	-0.75	-0.38	5.43	1.40	-0.27	0.5	9.06	0.46	0.51	1.47	1.54	6.99	1.76	-0.37	0.52	1.08	2.27	-0.19	-0.49	2

	scaled_avg_hotel_rooms	scaled_avg_checked_bags	scaled_cancellation_proportion	scaled_conversion_rate	scaled_weekend_trip_proportion
0	0.203140	0.718430	-0.252331	0.398117	-0.373519
1	-2.340000	-0.966199	-0.260000	-1.723496	-0.380000
2	0.171943	0.342979	3.462332	0.321010	0.004534
3	0.215061	0.108494	-0.238060	0.628589	2.301248
4	0.258524	-0.790872	-0.260000	-0.372912	-0.380000

	perk	user_count
0	Free Checked Bag	2299
1	Free Hotel Meal	1755
2	1-night free hotel with a flight	737
3	Exclusive Discounts	492
4	No Cancellation Fee	386

	scaled_avg_hotel_rooms	scaled_avg_checked_bags	scaled_cancellation_proportion	scaled_conversion_rate	scaled_weekend_trip_proportion
perk
1-night free hotel with a flight	0.215061	0.108494	-0.238060	0.628589	2.301248
Exclusive Discounts	-2.340000	-0.966199	-0.260000	-1.723496	-0.380000
Free Checked Bag	0.203140	0.718430	-0.252331	0.398117	-0.373519
Free Hotel Meal	0.258524	-0.790872	-0.260000	-0.372912	-0.380000
No Cancellation Fee	0.171943	0.342979	3.462332	0.321010	0.004534

	perk	dist_center
0	1-night free hotel with a flight	1.500694
1	Exclusive Discounts	0.331363
2	Free Checked Bag	1.209937
3	Free Hotel Meal	1.024512
4	No Cancellation Fee	2.027350

	scaled_avg_hotel_rooms	scaled_avg_checked_bags	scaled_cancellation_proportion	scaled_conversion_rate	scaled_weekend_trip_proportion	cluster_label	perk	dist_center
5494	-2.34	-1.11	-0.26	-1.82	-0.38	1	Exclusive Discounts	0.173181
1407	-2.34	-1.11	-0.26	-1.82	-0.38	1	Exclusive Discounts	0.173181
5771	-2.34	-1.11	-0.26	-1.82	-0.38	1	Exclusive Discounts	0.173181
1224	-2.34	-1.11	-0.26	-1.82	-0.38	1	Exclusive Discounts	0.173181
3225	-2.34	-1.11	-0.26	-1.82	-0.38	1	Exclusive Discounts	0.173181